OpenAI Drama - Act 2
It’s not dark yet, but it’s getting there
I wasn’t expecting to create another post so soon after Friday’s fun and games, but here we are.
The OpenAI implosion continued during and after the weekend. For a while it looked like Sam Altman and Greg Brockford might be coming back, with staff and investors unhappy at what had gone down. The staff threatened to resign en-masse, and set a deadline of 5pm on November 19th for the board to resign instead. This was agreed in principle but then stalled based on the makeup of the new board. Bret Taylor, former Co-CEO of Salesforce was one name both sides agreed on. Given that Bret was chairman of Twitter when Elon (eventually) bought it, he’d have to love drama to be up for this gig.
Everything then appears to have fallen apart, and when I got up at 6am on Monday for my usual dog walk, there was news of a CEO change at OpenAI. Not the one we were expecting though - the board had decided to stay and appoint Emmett Shear as their new CEO, after what appears to be minutes if not hours of executive search.
If the reason behind all this was that the board wanted to go slower, it looked like they’d got their man:
At this point, it really started to remind me of a Fun Boy Three single:
Then Satya Nadella posted that Sam Altman and Greg Brockman had joined Microsoft to head up a new advanced AI unit.
As well as a masterclass in leadership, this small piece of writing has some great edges:
Burying the lede/lead - starting with secondary details before sharing the essential information. We remain committed, look forward to, and Sam and Greg work for us now.
“We look forward to getting to know Emmett Shear” - clearly the short search and selection process didn’t involve Microsoft.
“We look forward to moving quickly” - which is what this is all about, apparently.
Pretty remarkable speed for an organisation the size of Microsoft, but Nadella clearly acting like the grown-up in the room.
Getting darker
The question was, how many OpenAI staff would go with them?
A few hours later we had an idea of the answer - all of them! A ‘resign or we walk’ letter was sent to the board, signed by 500 odd of the 770 staff.
Then things got really weird - in a Jeffrey Deaver-worthy twist, at number 12 on the list of people demanding the board resign, we find member of the board Ilya Sustkever, who posted on Twitter/X a seeming change of heart :
Another interesting piece of writing:
Regretting participation - word on the street is that he was the mover and shake, but presenting as more of a bystander.
The actions were the board’s, not his, even though he was and still is on the board
I’ll do anything to make it right - where was this sentiment at the end of last week?
Still, 500 out of 770 - a huge chunk of the company, right? Turns out that it was just those that were awake at the time:
How dark can it go?
In a few hours we went from OpenAI slowing down and maybe becoming irrelevant in a couple of years, to can they keep the lights on at the end of the week?
If nothing else, the OpenAI board look like they make terrible knee-jerk decisions and their comms are just appalling. The nihilist in me appreciates their seeming attempt to burn it all down, but for the life of me I can’t understand how they can continue to make decisions that backfire so quickly on them. Their product may be capable of great predictions, but it appears the humans aren’t. It turns out that companies that were worried about relying too much on OpenAI’s GPT models were right, but for the wrong reasons!
Through the Salesforce Lens
As always, I make a point of looking at this from a Salesforce angle. Right now I think, probably unintentionally, the fact that they rolled access to GPT features out slowly has turned out to be an inspired move:
Bring Your Own Model is probably looking good to a number of businesses also!
I did mention in my first post on this topic that it was never good to have a bus number of 1 for suppliers, but I thought that was a long term problem to solve rather than a short term scramble. I should point out that I don’t actually think OpenAI will close it’s doors - that would amount to negligence by the board in my view - but any company the size of Salesforce will be carrying out some serious risk assessment and mitigation work right now. OpenAI currently looks pretty unstable and the board are making a series of poor decisions. Any organisation that continues to rely on it based on the hope it will all sort itself out can look forward to answering a lot of uncomfortable questions if it does all go South.
I can’t imagine this is the end of it, as the board/staff stand off needs to be resolved somehow, probably by the board giving in to popular opinion and going. Hopefully it won’t lead to further days of drama, with more twists and turns than a Formula 1 race!